Time-Series Drone Onboard Multi-Modal Sensor Dataset

This dataset contains timeseries data from numerous drone flights. Each flight record has a unique identifier (uid) and a timestamp  indicating when the flight occurred. The drone’s position is represented by the coordinates (position_x, position_y, position_z) and  altitude. The orientation of the drone is represented by the quaternion (orientation_x, orientation_y, orientation_z, orientation_w). The  drone’s velocity and angular velocity are represented by (velocity_x, velocity_y, velocity_z) and (angular_x, angular_y, angular_z) respectively. The linear acceleration of the drone is represented by (linear_acceleration_x, linear_acceleration_y, linear_acceleration_z).

In addition to the above, the dataset also contains information about the battery voltage (battery_voltage) and current (battery_current) and  the payload attached. The payload information indicates if the drone operated with an embdded device attached (nvidia jetson), various sensors,  and a solid-state weather station (trisonica).

The dataset also includes annotations for the current state of the drone, including IDLE_HOVER, ASCEND, TURN, HMSL and  DESCEND. These states can be used for classification to identify the current state of the drone. Furthermore, the labeled dataset can be used for predicting the trajectory of the drone using multi-task learning.

For the annotation, we look at the change in position_x, position_y, position_z and yaw. Specifically, if the position_x,
position_y changes, it means that the drone moves in a horizontal straight line, if the position_z changes, it means that the drone performs ascending or descending (depends on whether it increases or decreases), if the yaw changes, it means that the drone performs a turn and finally if any of the above features do not change, it means the drone is in idle or hover mode.

In addition to the features already mentioned, this dataset also includes data from various sensors including a weather station and an Inertial Measurement Unit (IMU). The weather station provides information about the weather conditions during the flight. This information includes, wind speed, and wind angle. These weather variables could be important factors that could influence the flight of the drone and battery consumption. The IMU is a sensor that measures the drone’s acceleration, angular velocity, and magnetic field. The accelerometer provides information about the drone’s linear acceleration, while the gyroscope provides information about the drone’s angular velocity. The magnetometer measures the Earth’s magnetic field, which can be used to determine the drone’s orientation.

Field deployments were performed in order to collect empirical data using a specific type of drone, specifically a DJI Matrice 300 (M300).  The M300 is equipped with advanced sensors and flight control systems, which can provide high-precision flight data. The flights were designed  to cover a range of flight patterns, which include triangular flight patterns, square flight patterns, polygonal flight pattern,  and random flight patterns. These flight patterns were chosen to represent a variety of different flight scenarios that could be encountered  in real-world applications. The triangular flight pattern consists of the drone flying in a triangular path with a fixed altitude.  The square flight pattern involves the drone flying in a square path with a fixed altitude. The polygonal flight pattern consists of the drone  flying in a polygonal path with a fixed altitude, and the random flight pattern involves the drone flying in a random path with a fixed altitude. Overall,  this dataset contains a rich set of flight data that can be used for various research purposes, including developing and testing algorithms for  drone control, trajectory planning, and machine learning.